Run experiments in current process, without ExecutePreprocessor

[Jacob-Stevens-Haas]

This will allow us to escape the limitations on interprocess communication to a Jupyter notebook (I have to use a ZMQ socket or something? rather than pickle to a pipe?). It also makes it so we don't have to lookup/eval parameters twice: once in mitosis, once in the notebook itself.

Solution could take the form of building an html file from, say, jinja, or perhaps collecting output of experiment and setting the "output" entry of a cell (and metadata, as if it were run) without actually running the code in the cell.